SOLUTION · COUCHBASE CDC
SOLUTION · COUCHBASE CDC Couchbase CDC: change data capture and real-time data replication
Real-time Couchbase CDC from the native DCP stream, or from Eventing, so you pick the capture path your edition and cluster allow.
Gluesync by MOLO17 captures changes from Couchbase Server with a dedicated source agent: the Couchbase DCP agent reads the native Database Change Protocol stream, and the Couchbase Eventing agent uses the Eventing service. Gluesync then delivers those changes continuously to the databases, warehouses, lakes, and event streams your teams already use. Manage every pipeline from the Core Hub web UI, the Gluesync control plane.
WHO THIS IS FOR
Couchbase teams moving operational data into analytics and streaming platforms
- Couchbase DBAs who must approve CDC on a production bucket and want the exact role to grant:
data_dcp_readeron Enterprise Edition, orbucket_full_accesson Community Edition. The DCP path writes nothing to the source bucket. - Data platform leads feeding warehouses, lakes, SQL databases, or event streams from Couchbase, who want one Core Hub control plane for sources and targets rather than a Kafka mesh.
- Community Edition teams that need a native change stream without the Eventing service, which the DCP agent provides.
- Engineers who prototyped Couchbase CDC with a custom DCP client, Eventing functions, or a Kafka connector, and now own the on-call for a pipeline the business treats as infrastructure.
THE PROBLEM
Couchbase data that only moves in batches goes stale, and the DIY options carry their own cost
Couchbase is often the operational store, so scheduled exports, query-based polling, and reporting jobs leave analytics and downstream systems working from a copy that is already behind. Buyers searching for Couchbase CDC usually compare a Couchbase Kafka connector, a custom DCP client, Eventing functions they write and operate, and Couchbase's own XDCR for Couchbase-to-Couchbase replication. Each of those either adds a broker to run, puts the delivery code on your team, or serves only Couchbase destinations.
Gluesync addresses that pattern with per-agent CDC and two capture paths on Couchbase Server. The DCP agent and the Eventing agent share the same pipeline model and Core Hub, so you choose by edition, version, and whether the Eventing service runs in your cluster, rather than by rebuilding pipelines. Target agents write to the destination you choose.
HOW IT WORKS
How Gluesync does Couchbase CDC
Two native capture paths: DCP or Eventing
Both source agents read changes from Couchbase Server. They differ in where the change stream comes from, and that difference decides which edition and version you can use.
- DCP agent: reads the native Database Change Protocol (DCP) stream through
dcp-client. It does not need the Eventing service, deploys no Eventing functions, and writes nothing to the source bucket. DCP agent docs ↗ - Eventing agent: uses the Eventing service. Gluesync deploys and updates the Eventing functions it needs automatically, and those functions write event documents to the
gluesync-scopeandgluesync-collectioncollection. Eventing agent docs ↗
What the DBA will check
Prerequisites differ by path. These are the documented requirements, not a full runbook.
- DCP on Enterprise Edition: the
data_dcp_readerrole on the bucket, a scope, or a collection. It is a read-only role that can open DCP streams and read data. - DCP on Community Edition: use
bucket_full_accesson the source bucket, or a broader admin role. - Eventing: the Eventing service deployed in the cluster, at least 512 MB of RAM reserved for it before production, read and write access on the source bucket, and the
cluster-adminrole for automatic function deployment. - DCP ports: 8091 for management and bootstrap, 11210 for key-value (KV), and 11207 for KV over TLS. Use 18091 when TLS or Capella is enabled. The DCP path does not use the Eventing ports 8096 or 18096.
- Connections: TLS is a setting on the connection, with certificate upload. For Capella, enter the cluster hostname, for example
cb.XYZ.cloud.couchbase.com.
Versions and stream shape
The DCP agent's stream shape depends on the Couchbase version and the scopes and collections setting.
- Scopes and collections on: the server filters the stream by scope and collection. This requires Couchbase 7.0 or later.
- Scopes and collections off: the whole bucket is streamed. Couchbase 6.x is supported this way.
- Eventing: listed for Couchbase 7.0 and later on Enterprise Edition.
Architecture around Core Hub
Lightweight agents sit close to each system. Core Hub orchestrates pipelines, applies mapping and transformation, and serves the web UI and REST API. Pipelines group a source agent, target agents, and entities. Core Hub and agents deploy with Docker, Docker Compose, or Kubernetes.
CAPTURE OPTIONS
Two Couchbase source agents: how to choose
The choice comes down to edition and what your cluster runs. Use the DCP agent on Community Edition, or when the Eventing service is not part of your cluster. Use the Eventing agent on Enterprise Edition when the Eventing service already runs and you want its bundled functions. Both run under the same Core Hub pipelines and targets.
| Agent | Capture technique | Versions | Best for |
|---|---|---|---|
| Couchbase DCP agent ↗ | Native Database Change Protocol (DCP) stream, read through dcp-client, with no Eventing service | Couchbase 7.0 and later with scopes and collections, or 6.x without them; Community Edition and Enterprise Edition; Capella supported | Community Edition clusters, clusters without the Eventing service, and teams that do not want Eventing functions in the cluster. It writes nothing to the source bucket. |
| Couchbase Eventing agent ↗ | Couchbase Eventing service, with functions that Gluesync deploys and updates | Couchbase 7.0 and later; Enterprise Edition; Eventing service required in the cluster | Enterprise Edition clusters that already run the Eventing service and want the bundled functions, deployed and updated by Gluesync. |
TARGETS AND TOPOLOGIES
Keep Couchbase Server as the operational store while modernizing destinations
Use the integrations directory to pair Couchbase Server as a source with relational engines, NoSQL stores, cloud warehouses, object and lake storage, or event streams, subject to each agent's documented source and target role. Couchbase Server is also a target: the Couchbase target agent writes to Couchbase Server 7.0 and later, on-premises or in Capella.
Gluesync keeps pace with your change volume at any scale. MOLO17 Professional Services can size the deployment with your team.
Target agents write in optimized batches, never row by row, and switch to native bulk load for both snapshots and CDC on targets such as Snowflake, Google BigQuery, Amazon Redshift, Microsoft SQL Server, and PostgreSQL.
- Offload reads from Couchbase to a relational database or cache: see database offload
- Feed cloud warehouses or a data lake continuously: see warehouse sync and data lake
- Publish Couchbase changes to Apache Kafka or another event-streaming platform, alongside other targets
- Fan one Couchbase source out to several targets in the same pipeline design, under one Core Hub
FAIR, HIGH-LEVEL COMPARISON
Where Gluesync fits among Couchbase CDC approaches
| Approach | What buyers usually get | Where Gluesync fits |
|---|---|---|
| Couchbase Kafka connector with Kafka Connect | Changes flow through Kafka topics; you run Kafka and Kafka Connect, and you own connector upgrades and the consumers. | Gluesync delivers to Kafka as one target among many, without requiring a Kafka mesh for database-to-database or database-to-warehouse pipelines. Compare the Kafka and Debezium approach. |
| Couchbase XDCR | Native replication between Couchbase clusters, where the destination is another Couchbase cluster. | Gluesync targets are not limited to Couchbase. Relational engines, warehouses, lakes, and streams can be fed from the same Couchbase source under one Core Hub. |
| Eventing functions you write and operate | Full control inside the cluster; you own the function code, deployment, upgrades, and failure handling. | The Eventing agent deploys and updates the functions it needs. The DCP agent offers a path that needs no Eventing service at all. |
| Custom DCP client | Direct access to the native stream; you write the delivery, retry, and target logic for every destination. | The DCP agent reads the same native stream, with targets, entities, and Core Hub operations provided, and best-in-class MOLO17 enterprise support (rated 4.9/5 by customers) behind it. |
| Batch exports, scheduled queries, or timestamp polling | Simple to start; data is only as current as the last run, and the queries run against the operational cluster. | Continuous capture from the native change stream, with no scheduled export job to maintain. |
RELATED CONTENT
Couchbase CDC research and implementation detail
- Gluesync Couchbase DCP source agent: native CDC without Eventing
- Moneybase: analytics on live Couchbase data in Microsoft SQL Server
- Debezium alternative: managed CDC versus Kafka + Debezium
- Warehouse sync solution
- Couchbase agent overview ↗
- Change data capture (DCP) agent docs ↗
- Change data capture (Eventing) agent docs ↗
FAQ
Couchbase CDC questions
What is Couchbase CDC with Gluesync?
Change data capture from Couchbase Server through one of two source agents. The Couchbase DCP agent reads the native Database Change Protocol stream, and the Couchbase Eventing agent uses the Eventing service. Changes are delivered to the configured targets through Gluesync agents and Core Hub.
Which Couchbase agent should we choose, DCP or Eventing?
Choose the DCP agent for Community Edition, for clusters without the Eventing service, or when you do not want Eventing functions in the cluster. Choose the Eventing agent on Enterprise Edition when the Eventing service already runs and you want the bundled functions.
Which Couchbase versions and editions are supported?
The DCP agent supports Couchbase 7.0 and later with scopes and collections, and 6.x without them, on Community Edition and Enterprise Edition. The Eventing agent is listed for Couchbase 7.0 and later on Enterprise Edition.
What must the DBA grant for the DCP agent?
On Enterprise Edition, the Data DCP Reader role (data_dcp_reader) on the bucket, scope, or collection you stream. On Community Edition, bucket_full_access on the source bucket. The DCP path writes nothing to the source bucket.
What does the Eventing agent need in the cluster?
The Eventing service deployed, at least 512 MB of RAM reserved for it before production, and credentials with read and write access on the source bucket. Automatic function deployment also requires the cluster-admin role.
Does Couchbase CDC work on Capella?
Yes. The DCP agent is documented as verified with Capella, using the cluster hostname and port 18091, and the Couchbase target agent is documented for Capella clusters as well.
Can Gluesync write to Couchbase Server, not only read from it?
Yes. The Couchbase target agent writes to Couchbase Server 7.0 and later, on-premises or in Capella, using key-value operations, and uses N1QL for deletes of DataModeling entities.
CDC BY SOURCE DATABASE
Other sources Gluesync captures from
Evaluate Gluesync with your Couchbase change streams
Start a trial on your infrastructure, or talk to MOLO17 about the right Couchbase agent for your edition, version, prerequisites, and targets.